arXiv subjects

Qiancheng Zhang

Publications and source records attributed to Qiancheng Zhang.

4 recordsLinked to original sources

arXiv2026

FinBoardBench: Benchmarking Dynamic Wealth Management and Strategic Financial Reasoning of LLMs via Board Game Simulations

Recently, large language models (LLMs) have achieved superior performance in static financial reasoning and simple dynamic trading tasks. However, existing static financial benchmarks are insufficient to assess the dynamic wealth management and financial decision-making capabilities of LLMs in real-world environments. To bridge this gap, we present FinBoardBench, an evaluation suite based on three classic financial board games: Cashflow, Acquire, and Monopoly. FinBoardBench assesses a comprehensive set of financial skills, including personal cash flow management with debt balancing, corporate investment and acquisition forecasting, and competitive trade negotiations with asset auctions. Our experiments with 9 advanced LLMs reveal that while exhibiting basic long-term planning and investment logic, they fail to effectively leverage complex interactions for profit, and their strong static reasoning performance does not transform into successful dynamic decision-making. Notably, they tend to prioritize immediate asset acquisition over maintaining sufficient liquidity, making them vulnerable to financial crises triggered by random events. We hope that FinBoardBench can provide a valuable reference for more intelligent LLM-based decision-making systems in the future.

cs.CL↗

arXiv2026

What Factors Affect LLMs and RLLMs in Financial Question Answering?

Recently, large language models (LLMs) and reasoning large language models (RLLMs) have gained considerable attention from many researchers. RLLMs enhance the reasoning capabilities of LLMs through Long Chain-of-Thought (Long CoT) processes, significantly improving the performance of LLMs in addressing complex problems. However, there are few works that systematically explore what methods can fully unlock the performance of LLMs and RLLMs within the financial domain. To investigate the impact of various methods on LLMs and RLLMs, we utilize five LLMs and four RLLMs to assess the effects of prompting methods, agentic frameworks, and multilingual alignment methods on financial question-answering tasks. Our research findings indicate: (1) Current prompting methods and agent frameworks enhance the performance of LLMs in financial question answering by simulating Long CoT; (2) RLLMs possess inherent Long CoT capabilities, which limits the effectiveness of conventional methods in further enhancing their performance; (3) Current advanced multilingual alignment methods primarily improve the multilingual performance of LLMs by extending the reasoning length, which yields minimal benefits for RLLMs. Additionally, we discuss strategies for enhancing the performance of LLMs and RLLMs in financial question answering, which may serve as a inspiration for future improvements. We hope that this study can serve as an important reference for LLMs and RLLMs in the field of financial question answering.

cs.CL↗

arXiv2025

New constructions of $2$-to-$1$ mappings over $\gf_{2^n}$ and their applications to binary linear codes

The $2$-to-$1$ mapping over finite fields has a wide range of applications, including combinatorial mathematics and coding theory. Thus, constructions of $2$-to-$1$ mappings have attracted considerable attention recently. Based on summarizing the existing construction results of all $2$-to-$1$ mappings over finite fields with even characteristic, this article first applies the generalized switching method to the study of $2$-to-$1$ mappings, that is, to construct $2$-to-$1$ mappings over the finite field $\mathbb{F}_{q^l}$ with $F(x)=G(x)+{\rm Tr}_{q^l/q}(R(x))$, where $G$ is a monomial and $R$ is a monomial or binomial. Using the properties of Dickson polynomial theory and the complete characterization of low-degree equations, we construct a total of $16$ new classes of $2$-to-$1$ mappings, which are not QM-equivalent to any existing $2$-to-$1$ polynomials. Among these, $9$ classes are of the form $cx + {\rm Tr}_{q^l/q}(x^d)$, and $7$ classes have the form $cx + {\rm Tr}_{q^l/q}(x^{d_1} + x^{d_2})$. These new infinite classes explain most of numerical results by MAGMA under the conditions that $q=2^k$, $k>1$, $kl<14$ and $c \in \gf_{q^l}^*$. Finally, we construct some binary linear codes using the newly proposed $2$-to-$1$ mappings of the form $cx + {\rm Tr}_{q^l/q}(x^d)$. The weight distributions of these codes are also determined. Interestingly, our codes are self-orthogonal, minimal, and have few weights.

cs.IT↗

arXiv2025

Balanced Boolean functions with few-valued Walsh spectra parameterized by $P(x^2+x)$

Boolean functions with few-valued spectra have wide applications in cryptography, coding theory, sequence designs, etc. In this paper, we further study the parametric construction approach to obtain balanced Boolean functions using $2$-to-$1$ mappings of the form $P(x^2+x)$, where $P$ denotes carefully selected permutation polynomials. The key contributions of this work are twofold: (1) We establish a new family of four-valued spectrum Boolean functions. This family includes Boolean functions with good cryptographic properties, e.g., the same nonlinearity as semi-bent functions, the maximal algebraic degree, and the optimal algebraic immunity for dimensions $n \leq 14$. (2) We derive seven distinct classes of plateaued functions, including four infinite families of semi-bent functions and a class of near-bent functions.

cs.IT↗